Summary

In this chapter, we learned a few techniques for interpreting Machine Learning models. We saw that there are techniques that are specific to the model used: coefficients for linear models and variable importance for tree-based models. There are also some methods that are model-agnostic, such as variable importance via permutation.

All these techniques are global interpreters, which look at the entire dataset and analyze the overall contribution of each variable to predictions. We can use this information not only to identify which variables have the most impact on predictions but also to shortlist them. Rather than keeping all features available from a dataset, we can just keep the ones that have a stronger influence. This can significantly reduce the computation time for training a model or calculating predictions.

We also went through a local interpreter scenario with LIME, which analyzes a single observation. It helped us to better understand the decisions made by...